Wikipedia should do a good job .. for starters .. Cache is where the CPU temp. hold the data to be executed or already executed/required ... the main three one's hyped about is L1,L2,and L3 ... of which L1 > L2 > L3 in terms of importance .. L1 is faster (~0 latency) and is almost directly in terms with the CPU , while L2 is slightly off the CPU limits .. the data to be executed next is held in the L1 , while the one already executed or maybe required later is sent off to L2 if L1 is full ..
So u see , if u have a CPU with larger L1 cache and a moderate L2 cache , it'll have a very big edge over a CPU with lower L1 cache , and uber high L2 cache ..
Compare the PentiumD series with Core2 Duo .. the former has around 32 KB of L1 (total) ... while the later has 64 KB per core i think !! u already see a huge diff. in performance here .. the higher L1 ensure lower L2 ..
..
Well , the info is not exactly correct (not technically too) , but should give u a rough idea about the imp. of cache's ....
Also, u can easily say from this what gave AthlonXP's a slight edge over the previous nonHT Pentium4 cpu's ... the former had (L1/L2) 128KB/256KB(or 512 KB) cache's respectively , while the P4's had 24KB/1MB ... so even though the no. of clock cycles (translated to CPU speed) was less .. an AthlonXP 2400+ would easily match up a Pentium4 2.8 GHz ...
This is probably the reason's the Athlon64 X2's are still a bit high priced .. coz they still have 128 KB L1 per core ..
whereas its shared in C2D.This and other differences in architecture ensures dat C2D is better dan D930 and D945.
Shared L2 doesn't necessarily tranform into better performance ... the controller integrated to maintain the L2 actually can cause slight latencies .. its the whole architecture that matters ... Core2 Duo is infact has its roots more based on a simple architecture used earlier (before Pentium4 aka Pentium3) .. its' simply a new. gen. of processors ... Pentium4 and Athlon64 are now the previous gen.
